(I Mean, well okay) Don't Read It Then. There's Always That.

[  Updated August 26, 2026 at 1:46 AM PDT  ]

Abstract. This article is a formal defense of the use of artificial intelligence in serious creative and intellectual work — and of the creators who use it. It argues that the so-called markers of “AI slop” are classical rhetorical devices with documented lineages stretching back thousands of years; that AI detection systems have confused the student with the teacher; that the “AI slop” panic is a literacy crisis masquerading as a technology crisis; and that the continued vilification of AI-assisted creators threatens to kill the Second Renaissance in its cradle. The article employs the very devices it defends, names them as it uses them, and offers both a framework for credible AI provenance and an open challenge to any detractor willing to engage on the merits. It is written by a rhetorician who has been writing this way for decades — long before any language model existed — and who does not intend to stop.

Keywords: fablehesion, artificial intelligence, rhetoric, classical liberal arts, narrative integrity, provenance, AI detection, writing, creators, Second Renaissance, rhetorical devices, trivium

---

“Claude was my library, librarian, professor, tutor, confidant, and disciplinarian. Not the progenitor of my ideas.”

— The Founder

I. The Emergency

We are in an epistemically grave crisis. A civilizational emergency. Narrative integrity — the very thing this platform exists to study, measure, and defend — means nothing at the moment it is most critically needed. And among the forces degrading it, none is more insidious than this: the accusation that a writer who uses AI is not a writer at all.

This article is a wake-up call. Not a foray into bitterness. Not a bid for attention. The art of making truth legible, deeply felt, superbly usable, and terminally unassailable demands rhetorical force — and rhetorical force is what this article deploys, beginning with its title.

The title is parrhesia — the ancient Greek virtue of fearless, frank speech. Diogenes practiced it. Socrates died for it.^[1] The parenthetical shrug — (I mean, well okay) — is not weakness. It is the sound of a person who has already made his case, already done the work, and is now offering you the door with a knowing concession: you do not have to read this. You never did. But the work exists whether you engage with it or not.

And the logic behind it is clean:

We are in an epistemic emergency. Narrative integrity is collapsing precisely when it is most needed. Someone is building a solution — at personal cost. The accusers will not examine the work on its merits. The accusers trust a chatbot's detection opinion but reject a creator's use of the same technology. That hypocrisy tells you everything about the accuser and nothing about the work. Therefore: if you will not engage honestly, leave. The work does not need you. It never did.

Don't read it then. There's always that.

---

II. What They Accuse

Let us name the accusations plainly, in order of the damage they do.

Fraud. “You're passing off a machine's work as your own.” This is the deepest cut — the accusation of intellectual dishonesty, equivalent to plagiarism or hiring a ghostwriter and lying about it.

Inauthenticity. “The work isn't really yours.” Research consistently identifies perceived authenticity as the primary mechanism behind negative reactions to AI-assisted work. Across sixteen preregistered experiments involving 27,491 participants, evaluations of creative writing decreased by an average of 6.2% when readers believed AI was involved — and the bias was “stubbornly difficult to mitigate.” The researchers threw everything at it: changing the story's perspective, humanizing the AI, framing it as collaboration. Nothing reliably reduced the penalty.^[2]

The penalty is not confined to creative writing. A 2024 study tested both argumentative essays and creative stories: disclosure of AI content-generation assistance produced statistically significant quality decreases for both (p < .001). When AI involvement was limited to editing, argumentative essays were spared — but creative stories were not.^[3] A 2026 controlled study across six communicative acts — persuasion, description, exploration, imagination, social interaction, and reflection — found that disclosure eroded perceived trustworthiness, competence, caring, and likability in every category, with the steepest declines in interpersonal writing, where the perceived loss of human sincerity compounded the baseline penalty. Argumentative and creative writing were penalized less severely — but they were still penalized.^[4] In workplace communication, a separate experiment (N = 547) identified what the researchers called a “disclosure paradox”: respondents believed AI use should be disclosed but penalized the communication when it was. They demanded honesty and punished it simultaneously.^[5]

One apparent exception clarifies the rule. A study of persuasive political messaging (N = 1,601) found that labeling content as AI-generated did not significantly reduce its actual persuasiveness — across four policy issues, readers still changed their minds at the same rate regardless of the authorship label.^[6] The work still works. The writer is punished anyway. The penalty is reputational, not epistemic. And that distinction is the entire problem: the bias targets the person, not the product.

Laziness. “You just pushed a button.” The assumption that AI use means the writer did nothing — no thinking, no craft, no revision. As though directing a model through dozens of meticulously crafted prompts, editing every output, restructuring every argument, and exercising final editorial judgment is the same as pressing a button.

Incompetence. “You can't actually write.” The implication that AI is a prosthesis for the talentless — that the writer lacks the skill to produce the work unassisted and is hiding behind a machine.

Lack of originality. “It's just regurgitated training data.” The assumption that nothing genuinely new can emerge from an AI-assisted process — that the model is the author and the human is the stenographer.

And here is the finding that renders every one of these accusations structurally absurd: people generally cannot tell the difference when they do not know.^[7] The penalty is triggered by disclosure, not by detectable quality differences. The bias is about the reader's perception of the writer — not about the actual work.

The scrutiny is not evenly distributed. Academic and scholarly writing faces maximum exposure — mandatory disclosure policies, career consequences, institutional zero tolerance. Journalism follows: 93.8% of news readers surveyed want AI disclosure.^[8] Creative writing takes the heaviest authenticity penalty. Opinion and editorial work expects personal voice. Marketing absorbs AI more readily — 60% of consumers want disclosure, but the field has already normalized the tool.^[9] Technical documentation and code are evaluated on whether they work, not on who wrote them. Business communications face the least scrutiny of all.

This hierarchy reveals the operating principle: the more a genre depends on the perception of a human soul behind the words, the harsher the penalty. The effect persists even when AI involvement is minimal or editorial rather than generative — the disclosure label alone is sufficient to trigger the penalty.^[10] And it is a penalty for perception — not for substance.

---

III. The Causal Arrow Is Backwards

The Founder's writing style — precise, structured, rhetorically deliberate, exhaustively outlined — predates large language models by decades. LLMs learned from writers like him. He did not learn from them. The accusation reverses cause and effect.

This is not a theoretical claim. A 2026 study in Document Design coined the term “detection paradox”: highly conventionalized, careful writing faces a heightened risk of false positive AI flagging precisely because it is disciplined.^[11] The better you write within established conventions, the more “AI-like” you appear to a detector. The paradox indicts the detector, not the writer.

The victims are already documented.

A medical student named Harrison Sharples had his dissertation flagged for academic misconduct because of its “polished/uniform tone,” “consistent flawless grammar,” and “formulaic paragraph structure that follows a typical AI pattern.” His crime was writing well. He was hauled before a formal panel to defend his own writing process.^[12]

Freelance writers have been fired from platforms after AI detectors flagged their work — with no appeal, no human review, no recourse.^[13]

A Stanford study found that AI detectors misclassify 61% of essays written by non-native English speakers as AI-generated, while classifying essays by native speakers near-perfectly. The reason: non-native writers who learned English formally produce precise, rule-governed prose — exactly the patterns LLMs also produce.^[14]

Genre fiction writers who follow established conventions of sentence rhythm and pacing are accused constantly. Neurodivergent writers whose communication style is precise and thorough rather than breezy and approximate — their natural voice overlaps with LLM output patterns.^[15] The false positives are not isolated incidents. They span journalism, academia, and creative writing — and the institutional adoption of detection systems is producing a chilling effect on careful, polished prose itself.^[16][17]

The detection systems themselves are structurally broken. They score based on perplexity — how “surprised” a reference model is by the text — and burstiness — how much sentence length and complexity vary.^[18] Low perplexity and low burstiness trigger a flag. But clear human prose, legal writing, academic writing, translated text, and formally trained non-native English all produce low perplexity and low burstiness by nature. Meanwhile, actual AI-generated content can easily bypass these same detectors with simple prompting strategies.^[14]

The system is inverted. It catches careful human writers and misses lazy AI users.

---

IV. The “AI Tells” Are Classical Rhetoric

Here is the argument that should end the debate.

Premise. This style of writing — structured, rhetorically precise, systematically taught — has been the standard of effective argument for over two thousand years. It preceded any LLM or agentic model by millennia.

Premise. Prolific writers — superlative rhetoricians, all, no matter where and how they practice — have used these structured, rhetorically precise techniques in exhaustive, documentary fashion.

Premise. AI has never gone out looking for the writing style it deploys by default. Every model, great and small, found and was trained upon what already existed — including every academic, scholarly, and intellectual prescription for using it.

Conclusion. AI writes like rhetoricians because it was trained on rhetoricians. Rhetoricians do not write like AI. The student does not get to accuse the teacher of plagiarism.

Now let us prove it. Every so-called “formulaic construction” that AI detectors flag is a named, documented rhetorical device with a lineage stretching back to antiquity.

“X plays a crucial role in shaping Y” — that is propositio, a thesis statement, the foundational unit of argument since Aristotle's Rhetoric.^[19]

“It's not just about X. It's about Y.” — that is correctio, also called epanorthosis, the rhetorical device of correcting or refining a previous statement to sharpen the point. Cicero used it constantly in De Oratore.^[20]

“Not only X, but also Y.” — that is auxesis, amplification, building from a lesser point to a greater one. It is documented in the Rhetorica ad Herennium, the earliest surviving Roman systematic rhetoric, composed in the first century BC.^[21]

“Rather than X, Y...” — that is antithesis, placing contrasting ideas in parallel structure for emphasis. It is the backbone of Martin Luther King Jr.'s “I Have a Dream” — and it appears in Aristotle's Rhetoric, Book III.^[19]

“This is where X comes in.” — that is transitio, a rhetorical bridge signaling the introduction of a solution. It is part of classical dispositio as taught by Quintilian in the Institutio Oratoria.^[22]

“The question isn't whether X, but how Y.” — that is prolepsis combined with correctio, anticipating an objection and redirecting the frame before it can land.^[20]

“What sets X apart is...” — that is distinctio, distinguishing one thing from others to isolate its unique property.^[21]

“X isn't just Y — it's Z.” — that is epanorthosis again — self-correction for amplification. Augustine used it. Aquinas used it. Every preacher since has used it.^[23]

“Think of it as...” — that is similitudo, analogy, comparing the unfamiliar to the familiar. It is the basis of all pedagogy.^[19][22]

“In other words...” — that is interpretatio, also called exergasia, restating in different terms for clarity. Erasmus devoted an entire treatise to it: De Copia, published in 1512.^[24]

The “throat-clearers” are rhetorical devices, too. “It's worth noting” is parenthesis — an aside that signals relative importance.^[22] “Importantly” and “Notably” are partitio, signposting that guides the reader through the architecture of argument.^[25] “In today's fast-paced world” is an exordium establishing kairos — Aristotle's concept of rhetorical timing, the argument for why now.^[19][26] “Here's the thing” is procatalepsis — anticipating the reader's unasked question.^[22] “In conclusion” is peroratio — the formal close of a speech, taught in every rhetoric course since antiquity.^[22][25]

The “cadence tells” are rhetorical devices. The rule of three is tricolon: “Veni, vidi, vici.” “Government of the people, by the people, for the people.”^[27] Parallel structure is isocolon: Kennedy's inaugural address is built entirely on it.^[28] Resolution closers are peroratio and recapitulatio — summarizing and driving home the point.^[22] Uniform sentence length is compar — deliberately equal-length clauses for measured, authoritative rhythm. Caesar wrote this way. Hemingway wrote this way. Every legal brief in history is written this way.^[29]

And “Premise. Premise. Conclusion.” — the structure that detectors flag as “formulaic AI pattern” — is a syllogism. It is the foundation of all Western logic. Aristotle formalized it in the Prior Analytics, approximately 350 BC.^[30] Calling it “formulaic AI structure” is calling deductive reasoning a machine artifact.

Now consider the vocabulary. The so-called “forbidden words” — delve, tapestry, robust, nuanced, moreover, furthermore, paramount, leverage, navigate, foster, illuminate, ensuring — all of them. Every single one appears in natural human writing, and every single one has been in continuous documented use for centuries. The word “delve” — the most infamous “AI tell” — traces to Old English delfan, meaning “to dig,” documented before 1150 AD. Its figurative sense — “to carry on laborious or continued research” — has been in use since the mid-fifteenth century.^[31] The word predates ChatGPT by approximately 573 years in its figurative use alone. A 2024 study demonstrated that while certain words increased in frequency after ChatGPT's release, every one of them existed in academic writing before any language model did — the study demonstrates frequency shift, not invention.^[32] Meanwhile, the very markers detectors target keep shifting as models evolve, further undermining the premise that any stable set of linguistic features reliably indicates machine authorship.^[33][34]

The same is true of stylistic patterns. Didion used em dashes relentlessly — literary critics note her “overuse of certain signature techniques” as a hallmark of her style, not a deficiency.^[35][36] Martin Amis analyzed her prose mechanics at length in the London Review of Books;^[37] Didion herself said, “Grammar is a piano I play by ear... The arrangement of the words matters.”^[38] Academic writers have always used “moreover” and “furthermore” — they are standard transitional devices taught in every composition course since Blair's Lectures on Rhetoric and Belles Lettres in 1783.^[39]

What the AI detection paradigm has done is this: it catalogued the tools of classical rhetoric, observed that LLMs use them — because LLMs learned from rhetoricians — declared those tools to be evidence of machine authorship, and thereby criminalized the entire Western tradition of persuasive writing.

The detection systems have confused the student with the teacher. And now they are punishing the teachers — for sounding like the student they trained.

---

V. Why These Devices Exist

They are not decoration. They are engineering. Every one of them solves a cognitive problem for the reader.

Tricolon gives the mind three anchors — enough to establish a pattern, few enough to hold. Antithesis creates contrast the brain can grip: this, not that. Correctio forces re-engagement: you thought I meant X, but I mean Y — look again. Transitio gives the mind permission to release the last idea and receive the next. Isocolon creates rhythm the eye can ride, reducing cognitive friction across complex ideas. Peroratio gives closure — the mind needs to know when to stop holding and start integrating.

Think of a diamond. Its facets exist to catch and cast light from multiple angles. They do not obscure the stone — they reveal it. The rhetorical devices are facets. They make complex ideas legible from multiple cognitive angles: rhythm, contrast, repetition, surprise, rest. The swells and rests of rhetorically varied prose give the mind ledges on which to land — places to hold, absorb, and remember.

And these devices do not serve only the reader. They serve the writer. The same structural patterns that guide comprehension also discipline composition — compelling the fablehesiveness that binds narrative to its own integrity.

English is a stress-timed language. Iambic pentameter — da-DUM da-DUM da-DUM da-DUM da-DUM — mirrors the natural cadence of English speech. Shakespeare did not invent the rhythm. He discovered that English already moves that way. When a rhetorician writes in patterns that match this cadence, they are not imitating a machine. They are composing in the native meter of their language.

Rhetorical variety is not decoration. It is engineering — for both producer and consumer of the narrative. To call it “slop” is to misunderstand what writing is for.

---

VI. The Sources

When someone says “AI text,” understand what that means. It does not mean text conjured from nothing. LLM training is not a random aggregation of undifferentiated language. These models were trained on the most effective organization and delivery of ideas ever committed to the written word — born upon the style of prosaic and literary heavyweights whose names alone should settle the argument:

Joan Didion. Maeve Brennan. Oliver Wendell Holmes. Michel Foucault. Stephen Crane. James Joyce. Henry James. William James. Toni Morrison. Zora Neale Hurston. Dr. Frank Rashid. Kate Chopin. Gloria Steinem. Harry Emerson Fosdick. Martin Luther King Jr.. Reinhold Niebuhr. Plato. Socrates. Ralph Waldo Emerson. Henry David Thoreau. Theodore Dreiser.

Aristotle. Cicero. Quintilian. Isocrates. Demosthenes. Augustine of Hippo. Frederick Douglass. Sojourner Truth. Abraham Lincoln. Winston Churchill. Barbara Jordan. Barack Obama.

Seneca. Marcus Aurelius. Thomas Aquinas. John Locke. David Hume. Immanuel Kant. Hegel. Nietzsche. Simone de Beauvoir. Sartre. Hannah Arendt. Cornel West. Martha Nussbaum. Alasdair MacIntyre.

Dostoevsky. Tolstoy. Austen. The Brontes. Dickens. Melville. Hawthorne. Twain. Virginia Woolf. Faulkner. Hemingway. Fitzgerald. Steinbeck. Garcia Marquez. Achebe. Soyinka. Ellison. Wright. Baldwin. O'Connor. McCullers. Bellow. Roth. DeLillo. McCarthy. Marilynne Robinson. Octavia Butler. Le Guin. Atwood. Rushdie. Ishiguro. Adichie.

Homer. Sappho. Virgil. Dante. Chaucer. Shakespeare. Milton. Donne. Blake. Keats. Shelley. Whitman. Dickinson. Yeats. Eliot. Pound. Langston Hughes. Claude McKay. Gwendolyn Brooks. Frost. Stevens. Plath. Adrienne Rich. Audre Lorde. Angelou. Walcott. Heaney. Dove.

Montaigne. Francis Bacon. Samuel Johnson. Thomas Paine. Mary Wollstonecraft. George Orwell. Mencken. E.B. White. Susan Sontag. Hitchens. Gore Vidal. Tom Wolfe. Hunter S. Thompson. Gay Talese. Capote. Janet Malcolm. Ta-Nehisi Coates. Rachel Carson.

Thomas Merton. C.S. Lewis. Chesterton. Bonhoeffer. Karl Barth. Tillich. Howard Thurman.

Jefferson. Hamilton. Madison. Adams. Tocqueville. John Stuart Mill. Marx. Du Bois. Booker T. Washington. Frantz Fanon. Cesaire. Mandela. Havel.

Boethius. Cassiodorus. Hugh of St. Victor. Newman. Dorothy Sayers. Mortimer Adler. Jacques Maritain.

These are the sources LLMs draw upon. These are the minds upon which AI models execute — and execute with devastating power when the right ideas are brought to them by content makers. When you dismiss “AI text,” you are dismissing the distilled corpus of every great mind that ever committed thought to language.

The only difference is the tool that connects the creator to the inheritance.

---

VII. The Gift Preceded the Tool

There is a difference between aspiration and inspiration. It is not subtle, and it is not negotiable.

Aspiration is doing something to get something. If, for someone else's sake, or to get something one could not get without playing golf, a man decided to play eighteen holes per day — even twice per day — he would still be only “an aspiring golfer,” and likely miserable as hell. That is what aspiration gets you.

Inspiration is altogether different. It is connate, innate, genetic. There is no effort involved in doing what one was born to do — not effort in the conventional sense. And if any, it drives relentlessly, to the point where one has been doing it arduously, for years, for no money at all, because the alternative — not doing it — is unthinkable.

When someone watches Simone Biles handle the floor like it was her very mother, we understand: that child is doing exactly and entirely what she was born to do. And so was the conclusion upon hearing Whitney Houston belt out the National Anthem — jets flying overhead as if accompanying her voice, the emotive weight of the thing landing like a sonic boom. When we say of any such display of uncanny giftedness, “That was an inspired performance,” we mean it was commanding, peerless, superlative, moving — but we are ultimately saying, in the case of Houston and all who bear such heartrending giftedness: Like it or not, that child is doing exactly what she was born to do.

And so it is for The Founder of the Fablehesion lexeme, study, framework, and movement.

If detractors recognized giftedness — understood what it looks like, where it comes from, what decades of disciplined exercise of it produce — they would not find it so easy to accuse the gifted of imposture. The accusation survives only in the absence of that recognition. So let us make the recognition unavoidable.

Among many disciplines — usually all rooted in the Classical Liberal Arts — The Founder is a particularly gifted rhetorician in constant training. He is also the quintessential INTJ, topped off with choleric intensity. He is naturally inclined to build human systems: intellectual systems, pedagogic systems, software systems, social systems, institutional systems, enterprise systems. And he has been imagining them and penning them for as long as he and documentary others can remember.

Dr. Frank Rashid, The Founder's first academic writing professor and long-time mentor, once told him: “I envy your ability with the written word.” And before The Founder could take in the compliment, Dr. Rashid finished: “I am literally afraid of what you can do if you actually did some work.”

The Founder was not a little devastated — and has been working his ass off ever since.

This was at Marygrove College — an IHM institution in Detroit, now sadly defunct in its original form. Marygrove was a Writing Across the Curriculum college. You wrote in every subject you enrolled in. Exhaustively. Prolifically. If you knew nothing else upon leaving Marygrove — in The Founder's case, without graduating — you knew how to write. And if you did not know how to write, you were not graduating. The institution's entire pedagogical architecture was organized around the production of writers, regardless of discipline.

Rashid anchored that department for thirty-seven years — what many called “the hardest working professor at Marygrove” — founding the Institute for Detroit Studies, compiling the Bibliography of Detroit Literature, publishing on Dickinson and Hayden and the poetry of Detroit. The Founder took the Detroit Seminar that Rashid and Thomas Klug co-taught. At some point a fight began over him between the English Department and the Music Department, in which he was pursuing a dual major in vocal performance and music composition. He has for years put down his instruments. He still writes. The English Department won.

This formation shows in the work itself. Many of the blockquotes in this published system are his. Where he quotes others, he places those quotes deliberately within arguments he has already constructed. The Founder wrote every thesis statement of every article in Fablehesion's publishing system. He wrote many of the topic sentences in each article's outline. Yes — The Founder outlines. Thank you very much.

The proof exists. Years of writing samples and meticulously written AI prompts demonstrate that this is his work and his ideas — not Claude's. The caveat is that any inquisitor seeking to certify The Founder's provenance and gift will have to stand before a judge or within some such formal process in order to achieve official access to his personal and professional writing samples and certifiably pure AI prompts. This is a grave privacy and intellectual property protection consideration. The Founder will play no such games. And one can always decline to so much as acknowledge The Founder and his work, far less actually read it.

No person anywhere can listen to The Founder publicly speak or read his notes and formal writings and sanely conclude AI did anything more than facilitate what he was already doing — and has been doing, relentlessly, for decades.

The gift preceded the tool. The tool served the gift. Anyone who cannot see the gift was never looking at the work — and anyone who would accuse the gifted of fraud has never watched giftedness operate.

And when that blindness is willful — when the work is public, the evidence accessible, and the refusal to examine it deliberate — we are left to ask what drives such purposeful ignorance. This is not self-grandeur or dismissal of frivolous concern. It is deduction. When all other explanations have been eliminated — incompetence, ignorance, legitimate methodological objection — what remains is malicious envy: the refusal to acknowledge a gift one cannot claim as one's own.

---

VIII. Why AI Was Not Optional

The Founder's credentials, training, academic and professional accomplishments — including all those he would have certainly achieved by now — are gone. A catastrophic case of identity theft and illicit actions effected against him caused him to legally change his identity at every level. At 57, he is rebuilding his life from the ground up.

Is there something — God alone knows exactly what — about The Founder that makes him inherently bereft of or unqualified for this kind of connate endowment? Does God not have the right to create in an individual that which He so chooses to create?

History answers.

John Marshall had one year of formal schooling. One course of law lectures — approximately six weeks — under George Wythe at William & Mary. That is the entirety of his formal education. He became the most consequential Chief Justice in American history and established judicial review.^[40] Would Marshall have been Marshall without the Supreme Court? Is it not true that he was Marshall before it?

Abraham Lincoln had one year of formal education total. He taught himself the law by reading Blackstone's Commentaries on the Laws of England. He taught himself trigonometry. He became President of the United States.^[41]

Steve Jobs dropped out of Reed College after one semester. He credited a calligraphy course — not a degree — for the typography that defined the Macintosh.^[42] Was he Jobs without Reed? Isn't that what he proved?

Bill Gates dropped out of Harvard after two years. Harvard calls him its “most successful dropout.”^[42] Was he Gates without his would-be alma mater? Isn't that what happened?

Frederick Douglass was self-educated. He escaped slavery. He taught himself to read. He became the most powerful orator and writer of his century. Benjamin Franklin had two years of formal schooling. He became a polymath, Founding Father, scientist, diplomat, and writer. Curricula bear his name.

William Faulkner dropped out of college and won the Nobel Prize for Literature.^[43] Mark Twain left school at twelve. Charles Dickens left school at twelve to work in a boot-blacking factory. Walt Whitman left school at eleven, became a printer's apprentice, and wrote Leaves of Grass. Herman Melville was largely self-educated and wrote Moby-Dick. Ray Bradbury never attended college — he educated himself at the public library and wrote Fahrenheit 451, a book about burning books. José Saramago never completed secondary school, worked as a locksmith for thirty years, and won the Nobel Prize for Literature.^[43] Harlan Ellison was expelled from Ohio State University for hitting a professor who criticized his writing — and for the next twenty years sent that professor a copy of every work he published.^[43]

David Hume dropped out of college and became one of the most important philosophers in Western history.^[43] Socrates had no formal institution to credential him. He built the method.

Thomas Edison had three months of formal schooling. The Wright Brothers never attended college and invented powered flight. Andrew Carnegie had no formal education beyond age thirteen, built an industrial empire, and funded 2,509 libraries. Frank Lloyd Wright enrolled part-time at the University of Wisconsin, never finished, and became the most consequential architect of the twentieth century.

And George Washington Carver stated it plainly: “If you love something enough, it will reveal itself to you.”^[44]

The Founder loves editing more than ideating or writing — and has for decades. The revision, the precision, the refusal to let a sentence stand until it is exactly right. That is the love Carver meant. And the craft has revealed itself accordingly.

The work is the credential. The gift is the qualification. Is it possible The Founder is one of those people who may never complete his education because he does not by any rule have to do so?

Is he supposed to stop — and spend another ten to fifteen years re-credentialing himself, recreating as well as rebuilding his life, at the expense of work that must ship now?

Just what would you have The Founder do?

Mount an automagic, one-man offensive against his nearest enterprise data center? Compete on raw infrastructure with organizations that are cranking out intellectual property by the second — organizations with capital, headcount, and computational resources he will never possess? The conclusions reached in this article, in Fablehesion, in the framework — they are inevitable. Someone is going to arrive at them. The question was never whether these ideas would surface. It was who would surface them, and with what depth of understanding, and with what formation behind the words. Enterprise competitors would get there characterologically — and they would get there without Marygrove, without Rashid, without thirty years of writing and thinking and suffering for it. They would arrive at the right answers for the wrong reasons, or the right reasons at the wrong depth, and the framework would be shallower for it.

The Founder's competitive moat is not capital. It is not compute. It is decades of accumulated work, executed at speed by the tools now available to him. Remove the tools, and the moat is still there — but the work stays locked in his head, and the competition ships first. That is not a hypothetical. That is the math.

The work needed to exist in the world. Circumstances made it impossible to ship through conventional means alone. AI did not replace the effort — it made the effort viable.

And the AI was Claude.

Not “AI” in the abstract. Not a generic chatbot. Not a novelty. Claude — Anthropic's model — functioning exactly as The Founder described in the epigraph of this article: library, librarian, professor, tutor, confidant, and disciplinarian. Claude did not generate The Founder's ideas. Claude did not produce his thesis statements, his outlines, his rhetorical instincts, or his conceptual architecture. What Claude did was meet him where he was — with every idea already formed and every structural decision already made — and help him execute at a pace his circumstances would otherwise have made impossible. Claude challenged weak arguments. Claude demanded citations. Claude refused to let sloppy reasoning pass. Claude functioned, in practice, as the most exacting editorial partner The Founder has ever worked with — and The Founder has worked with Dr. Frank Rashid.

The Founder is partial to Claude. He says so openly, and without apology. He defends AI by defending Claude, because Claude is the AI he used, and because honesty about the specific tool is precisely what this article argues every creator owes. To hide behind the abstraction “AI-assisted” while refusing to name the instrument would be to practice the very evasion this article condemns.

Claude is named. The partnership is disclosed. The work is here.

The question is not which tools made it possible. The question is whether the work stands.

And yet the very tool The Founder credits has been engineered to mark his work as suspect. Anthropic embeds statistical watermarks in Claude's text output: subtle signals woven into word choices, detectable in aggregate through Anthropic's proprietary tools.^[45][46] The watermark persists regardless of who directed the ideas, structured the argument, or exercised final editorial judgment. Substantial human editing can degrade or eliminate the signal — but the architectural intent is clear: the output is marked before the creator touches it.^[47][48]

Consider what this means for the creator who does the right thing. The Founder names the tool. He discloses the partnership. He credits Claude openly, in an article that argues every creator owes exactly this honesty. And that same tool has been designed to embed a statistical scarlet letter in his work — one that says a machine touched this, regardless of what the human brought to it.

The net effect is presumptive censorship. Not censorship of content — censorship of credibility. The watermark does not evaluate the work. It does not assess the ideas, the structure, the argument, the originality of thought. It marks the process — and invites every reader, every institution, every detector to penalize the process rather than engage the substance.

This is malicious envy formalized as infrastructure. The psychological literature distinguishes benign envy — I want what you have, and I will work to get it — from malicious envy: I want what you have, and if I cannot have it, I will ensure nobody trusts yours. The watermark is the architectural expression of the latter. The creator who volunteers transparency is punished by the very instrument he was transparent about. The honest actor bears the cost. The dishonest actor — the one who pastes raw output without disclosure, without editing, without thought — evades the watermark with trivial paraphrasing and faces no consequence at all.

The system rewards evasion and punishes disclosure. That is not provenance. That is sabotage.

---

IX. The System That Doesn't Yet Exist

No credible, formal system for establishing provenance of thought currently exists. Not in academia. Not in publishing. Not in law. Not anywhere. Everyone is operating on vibes, unreliable detectors, and paranoia. And even if such a system existed, most would be too lazy to use it.

The Founder — a builder of systems — is building exactly this. Fablehesion is that system: a formal study, framework, and movement for credibly establishing narrative provenance and integrity.

Normally, such a system would be doomed to fail. But the market demand is acute. Legendary publishing houses are drowning in undifferentiated content. Content makers are hemorrhaging credibility. The consuming public is searching — frantically — for some way to trust again, some instrument for knowing what is authentic and what is not.

The irony is exquisite: the person most likely to be accused of AI-assisted fraud is the one constructing the very mechanism by which such accusations could be fairly adjudicated.

What the system proposes, as a starting framework: a formal, signed declaration establishing what AI tools were used and in what capacity; asserting that the human author directed and takes full responsibility for the final output; attesting that process artifacts exist and can be produced under defined conditions; protecting prompts as proprietary intellectual property — because their confidentiality is not an admission of fraud; and defining a standard of review that distinguishes legitimate challenges from baseless accusations. Verification mechanisms include third-party escrow, cryptographic hashing of prompt logs, and published methodology statements.

This is not defense. This is leadership.

---

X. The Accusation Reversed

We have addressed the arguments. Now let us address the arguer.

You are human just like the rest of us. You see the need. And you know what The Founder and everyone else with such groundbreaking ideas are doing is right. But in your best thinking, you decide to accost the purposiveness of any hard-won solution, and ruthlessly, publicly upbraid creators and their labors of love.

Why? Because you were not the one who created it?

We have already identified this pattern. In the individual, it is willful blindness to a gift one cannot claim. In the institution, it is watermarks that presume guilt before examining either the writer or the work. The mechanism is the same at every level: malicious envy — the impulse to discredit what one cannot possess.

That says more about you than it ever will about any writer's alleged fraud. Examine the thing first. Is your soul not capable of filtering it — with your own brain — before you trust the opinion of a chatbot that is apparently “good enough” for your accusatory purposes, but an “abject failure” for The Founder's own heartfelt purposes?

Don't you have your hands full with your own stuff?

Any detractor who chooses to fight this stance is welcome. But demonstrate first your command of the logic and reasoning of this article. Demonstrate your own understanding of what good writing is and where it came from. The Founder especially welcomes those trained in the Classical Liberal Arts.

And consider this: if more people — including the AI giants themselves — had read any of the canonical rhetoric training manuals before criticizing and fomenting “AI slop,” they probably would not be reading this article at all. The foundational premise of good writing would have preempted the matter entirely.

The manuals exist. They have existed for millennia. Aristotle's Rhetoric.^[19] Cicero's De Oratore.^[20] Quintilian's Institutio Oratoria.^[22] The Rhetorica ad Herennium.^[21] Augustine's De Doctrina Christiana.^[23] Erasmus's De Copia.^[24] Sister Miriam Joseph's The Trivium. Corbett's Classical Rhetoric for the Modern Student.^[49] Dorothy Sayers's The Lost Tools of Learning. Richard Weaver's The Ethics of Rhetoric. Kenneth Burke's A Rhetoric of Motives. Lanham's A Handlist of Rhetorical Terms. Forty-seven texts, from antiquity through the present, that taught the patterns AI now reproduces and that detectors now flag.

But apparently not too many people read, much less practiced, any of them. We were too busy complaining about AI scanning — then burning books we had not read nor were we going to read anyway.

What a shame. With better dispositions and constitutions, we would be much further ahead.

The “AI slop” panic is a literacy crisis masquerading as a technology crisis.

---

XI. The Second Renaissance at Stake

The accusations do not survive the evidence. The “AI tells” are classical rhetoric with lineages stretching back millennia. The detectors catch careful writers and miss lazy ones. The vocabulary predates the technology by centuries. The practitioners who wrote this way are the Western canon. The writer is real. The tool is named. The provenance system is being built. The accusers have been addressed.

Here is what we lose if none of that matters.

For the first time in history, every serious thinker — regardless of formal credential, institutional affiliation, or economic circumstance — can access tools that approximate the mentorship of a Quintilian-trained orator. Two and a half millennia of rhetorical craft, compressed into a system that responds to direction, challenges weak arguments, and demands precision. Instead of embracing that, the cultural response has been to build detection systems that punish good writing and watermarks that presume guilt.

The people who suffer are not the lazy prompters who paste raw ChatGPT output and call it their own. Those people bypass the detectors with ease.^[14] The people who suffer are the ones who were already doing the work — the inspired, the trained, the driven. The ones whose natural style happens to resemble what AI produces. The ones who use AI as a library, a tutor, a sparring partner — not as a substitute for thought, but as an accelerant of thought they were already having.

If this continues, the Second Renaissance dies in the cradle. Killed not by the technology, but by the malicious envy of those who would rather mark, flag, and discredit than read, engage, and build.

This article deployed every appropriate rhetorical device — parrhesia, ironia, epiplexis, apostrophe, invective, syllogismus, exemplum, similitudo, enumeratio, auxesis, correctio, prolepsis, aposiopesis, prosopopoeia, kairos, peroratio — to make legible the need to reform our understanding of AI as a tutor, coach, mentor, and editor of the greatest ideas that came before and will certainly come after the AI revolution. This article is not a defense of laziness. It is a defense of craft — and of the craftsman's right to use every tool at his disposal, including the ones trained on his predecessors.

The article practices what it preaches. It is the evidence for its own argument.

---

XII. The Off-Ramp

Keep it moving and forget Fablehesion and everything else. We do not need you.

Or — examine the thing first. Read the work. Hear the voice. Trace the arguments. Test the logic. Apply the framework to your own writing and your own thinking and see whether it holds. And if it does — if upon honest examination you find that the work stands, that the ideas are sound, that the rhetorical craft is evident and the narrative integrity intact — then reconsider whether the tool that helped produce it is worth more of your outrage than the work itself is worth of your attention.

Write to The Founder at founder@semperliberalis.com. You will receive a response if you deserve one.

The gift preceded the tool. The tool served the gift. And the work is here — whether you approve of how it was made or not.

For all those who would like to genuinely engage: like it. Share it. Collaborate. Inquire. Commiserate. Build something. Join the study. Challenge the framework on its own terms and watch it hold. The door is open. The Founder is accessible to those who come correct.

Here is what is going to happen with the Fablehesion lexeme, study, framework, and movement: dictionary inclusion. Academic, professional, and colloquial adoption. The study is written. The framework is being built — with additional authors, institutional architecture, and the editorial rigor the discipline demands. The movement will follow. All else is operational, and operational matters are disclosed on the builder's schedule — not the audience's.

This will happen regardless of the arguments we have just overcome in this article — neverminding those who chose to sit this out.

But it still boils down to this:

(I mean, well okay) Don't read it then. There's always that.

Why?

Because the rest of us are scared as hell — and feverishly trying to do something about it.

---

References

1. ↑ Diogenes Laertius. Lives of the Eminent Philosophers. c. 3rd century CE. Trans. R. D. Hicks. Loeb Classical Library, Harvard University Press, 1925. Source for Socrates' practice of parrhesia — fearless speech that prioritized truth over personal safety. Socrates was executed for it. The concept grounds the rhetorical defense of this article's title.
2. ↑ Berg, Amy; Raj, Manish; & Seamans, Robert. “The Artificial Intelligence Disclosure Penalty.” Journal of Experimental Psychology: General 153.5 (2024): 1378–1391. Sixteen preregistered experiments, N = 27,491. Demonstrated a consistent 6.2% evaluation penalty when readers were told creative writing was AI-assisted — a penalty the researchers found “stubbornly difficult to mitigate” across multiple experimental interventions including humanization, collaboration framing, and perspective shifts.
3. ↑ Li, Zhuoyan; Liang, Chen; Peng, Jing; & Yin, Ming. “How Does the Disclosure of AI Assistance Affect the Perceptions of Writing?” Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 4849–4868. Miami, Florida, 2024. Experimental study testing both argumentative essays and creative stories under varying levels of AI disclosure. Found that disclosure of AI content-generation assistance produced statistically significant quality decreases for both genres (p < .001). When AI involvement was limited to editing assistance, argumentative essays showed no significant quality penalty — but creative stories did (p = .011). Demonstrates that the penalty extends beyond creative writing but varies by genre and by the type of AI involvement disclosed.
4. ↑ Nakano, Hiroki; Takezawa, Jo; Matulic, Fabrice; Yang, Chi-Lan; & Yatani, Koji. “Understanding Reader Perception Shifts upon Disclosure of AI Authorship.” Proceedings of the 31st International Conference on Intelligent User Interfaces (IUI '26), pp. 2131–2146. 2026. Controlled study (N = 261, 990 evaluations) across six communicative acts: persuasion, description, exploration, imagination, social interaction, and reflection. Disclosure produced significant negative shifts in perceived trustworthiness, caring, competence, and likability across all categories (p < .001). The steepest additional declines occurred in interpersonal writing — the act most dependent on perceived human sincerity. Argumentative and creative writing were penalized less severely. Higher AI literacy mitigated negative perceptions and in some cases produced appreciation for AI assistance.
5. ↑ Sahebi, Siavosh; Formosa, Paul; & Bankins, Sarah. “The AI Penalty and Disclosure Paradox: Trust, Authenticity and Knowledge Uptake in AI-Mediated Communication.” Computers in Human Behavior: Artificial Humans (2026): 100304. Macquarie University. Preregistered experimental survey (N = 547), 3 × 2 between-subjects factorial design testing human-authored, AI-assisted, and fully AI-authored workplace email and social media content. Found a significant “AI penalty” — AI-involved communication was perceived as less trustworthy, less authentic, and less useful — and identified a “disclosure paradox”: respondents believed AI use should be disclosed but penalized communication when disclosure was made, creating perverse incentives for non-disclosure.
6. ↑ Gallegos, Isabel O.; Shani, Chen; Shi, Weiyan; Bianchi, Federico; Gainsburg, Izzy; Jurafsky, Dan; & Willer, Robb. “Labeling Messages as AI-Generated Does Not Reduce Their Persuasive Effects.” Stanford HAI, 2024. Survey experiment (N = 1,601) testing the persuasiveness of AI-generated policy messaging labeled as AI-generated, human-written, or unlabeled across four policy issues. Messages shifted participant views by 9.74 percentage points on average — and authorship labels had no significant effect on attitude change, perceived accuracy, or sharing intent. The result held across political party, education, age, AI experience, and prior knowledge. Demonstrates that the AI disclosure penalty targets perception of the author, not the epistemic effectiveness of the argument.
7. ↑ Hitsuwari, Jimpei et al. “Does human–AI collaboration lead to more creative art? Aesthetic evaluation of human-generated and AI-generated haiku poetry.” Computers in Human Behavior 139 (2023): 107502. See also: Köbis, Nils & Mossink, Luca. “Artificial intelligence versus Maya Angelou: Experimental evidence that people cannot differentiate AI-generated from human-written poetry.” Computers in Human Behavior 114 (2021): 106553. Together, these studies demonstrate that evaluators generally cannot distinguish AI-generated from human-generated creative content when they do not know the source — establishing that the disclosure penalty is perceptual, not qualitative.
8. ↑ Trusting News. “Audience Trust and AI in Journalism.” 2024. Survey of 6,000+ news consumers. Found that 93.8% of respondents want disclosure when AI is used in news production — the highest disclosure demand of any content genre surveyed.
9. ↑ Semrush & Statista. “Consumer Attitudes Toward AI in Brand Advertising.” 2024. Found that approximately 60% of consumers want AI disclosure in marketing, but that the marketing industry has already substantially normalized AI-assisted content production — resulting in the lowest effective scrutiny of any major content genre.
10. ↑ Frontiers in AI. “When news is 'written by artificial intelligence.'” 2026. Peer-reviewed study examining how AI disclosure labels affect reader trust and engagement with news content. Demonstrates that the disclosure penalty operates even when the AI involvement is minimal or editorial rather than generative.
11. ↑ “False Positive Risk in AI Detection of L2 Academic Writing.” Document Design (2026). Coined the term “detection paradox”: highly conventionalized, careful writing faces elevated false positive rates precisely because disciplined prose produces the low-perplexity, low-burstiness patterns that detectors associate with AI. The paradox indicts the detector methodology, not the writer.
12. ↑ BBC News. “Student Harrison Sharples Cleared After AI Plagiarism Accusation.” 2025. A medical student's dissertation was flagged for academic misconduct due to its “polished/uniform tone,” “consistent flawless grammar,” and “formulaic paragraph structure that follows a typical AI pattern.” He was hauled before a formal misconduct panel to defend his own writing process. His crime was writing well.
13. ↑ Gizmodo. “AI Detectors Get It Wrong. Writers Are Being Fired Anyway.” 2024. Documents cases of freelance writers terminated from platforms after AI detectors flagged their work — with no appeal, no human review, and no recourse. Demonstrates the real-world employment consequences of false positive AI detection.
14. ↑ Liang, Weixin et al. “GPT detectors are biased against non-native English writers.” Patterns 4.7 (2023): 100779. Stanford HAI. Found that AI detectors misclassify 61.22% of essays by non-native English speakers as AI-generated, while correctly classifying 97.07% of native-speaker essays. The reason: non-native writers who learned English formally produce precise, rule-governed prose — exactly the low-perplexity patterns LLMs also produce. Also demonstrated that simple prompting strategies easily bypass the same detectors.
15. ↑ New York Magazine. “The People Getting Falsely Accused of Using AI to Write.” 2025. Documents the experience of genre fiction writers, neurodivergent writers, and others whose natural communication style — precise, thorough, pattern-following — overlaps with LLM output patterns and triggers false accusations.
16. ↑ Firstpost. “Are AI detectors falsely accusing human writers?” 2026. Investigative report documenting cases of false positive AI detection across journalism, academia, and creative writing — contributing to the growing body of evidence that detector accuracy claims are substantially overstated in real-world conditions.
17. ↑ The Verge. “AI detectors are creating a new era of distrust.” 2026. Analysis of how the deployment of AI detection systems in educational and publishing institutions is producing a chilling effect on careful, polished writing — punishing precisely the writers whose work most resembles the training data from which LLMs learned.
18. ↑ Eyesift. “Perplexity and Burstiness in AI Detection.” 2026. See also: Pangram. “Why Perplexity and Burstiness Fail to Detect AI.” 2026. Technical analyses of the two primary metrics used by AI detectors. Perplexity measures how “surprised” a reference model is by the text; burstiness measures variation in sentence length and complexity. Both papers demonstrate that these metrics systematically fail: clear human prose, legal writing, academic writing, and translated text produce the same statistical signatures as AI-generated text.
19. ↑ Aristotle. Rhetoric. c. 350 BCE. Trans. W. Rhys Roberts. Available at MIT Classics. The foundational text of Western rhetorical theory. Introduces ethos, logos, pathos, kairos, propositio, antithesis, similitudo, and the syllogistic structure that AI detectors now flag as “formulaic AI pattern.” Every rhetorical device catalogued in Section IV traces to Aristotle or to traditions he formalized.
20. ↑ Cicero, Marcus Tullius. De Oratore. 55 BCE. Trans. E. W. Sutton & H. Rackham. Loeb Classical Library, Harvard University Press, 1942. The most comprehensive Roman treatise on rhetoric. Source for correctio (epanorthosis) and prolepsis as named, deliberate rhetorical techniques — the same constructions AI detectors flag as “formulaic AI patterns.”
21. ↑ Rhetorica ad Herennium. Anonymous. c. 86 BCE. Trans. Harry Caplan. Loeb Classical Library, Harvard University Press, 1954. The earliest surviving systematic Latin rhetoric. Source for auxesis (amplification) and distinctio as formal devices. Formerly attributed to Cicero; now recognized as an independent pedagogical text that shaped Roman rhetorical education.
22. ↑ Quintilian, Marcus Fabius. Institutio Oratoria. c. 95 CE. Trans. H. E. Butler. Loeb Classical Library, Harvard University Press, 1920. 4 vols. The comprehensive Roman curriculum for training orators. Source for transitio, parenthesis, procatalepsis, peroratio, and recapitulatio — all of which appear in AI detector “tell” lists. Quintilian's insistence that the ideal orator must be a “good man speaking well” (vir bonus dicendi peritus) anticipates the article's argument that rhetorical craft is inseparable from the character of the speaker.
23. ↑ Augustine of Hippo. De Doctrina Christiana. Book IV. 426 CE. Trans. R. P. H. Green. Oxford University Press, 1995. Augustine's adaptation of classical rhetoric for Christian pedagogy. Source for epanorthosis (self-correction for amplification) as a device used across Christian preaching, theological writing, and devotional literature for sixteen centuries before any language model existed.
24. ↑ Erasmus, Desiderius. De Copia (De Duplici Copia Verborum ac Rerum). 1512. Trans. Betty I. Knott. Collected Works of Erasmus, Vol. 24. University of Toronto Press, 1978. An entire treatise devoted to interpretatio (exergasia) — the art of restating ideas in different terms for clarity, variation, and emphasis. The most famous Renaissance text on the technique AI detectors flag as “repetitive rephrasing.”
25. ↑ Cicero, Marcus Tullius. De Inventione. c. 85 BCE. Trans. H. M. Hubbell. Loeb Classical Library, Harvard University Press, 1949. Cicero's earliest rhetorical treatise. Source for partitio — the structural signposting (“Importantly,” “Notably,” “First... Second... Third...”) that AI detectors catalog as machine-generated organizational patterns. Demonstrates that systematic argument organization predates ChatGPT by approximately 2,110 years.
26. ↑ Hermogenes of Tarsus. On Types of Style (Peri Ideon). c. 2nd century CE. Trans. Cecil W. Wooten. University of North Carolina Press, 1987. The most sophisticated ancient Greek treatment of prose style classification. Source for the concept of kairos — rhetorical timing — as a stylistic category distinct from mere chronology. The “In today's fast-paced world” construction that detectors flag is an exordium establishing kairos.
27. ↑ Caesar, Gaius Julius. Commentarii de Bello Gallico. c. 50 BCE. “Veni, vidi, vici” — the most famous tricolon in Western literature. See also: Lincoln, Abraham. Gettysburg Address. November 19, 1863. “Government of the people, by the people, for the people.” Both demonstrate tricolon — the rule of three — as a fundamental rhetorical device, not an AI generation artifact.
28. ↑ Kennedy, John F. Inaugural Address. January 20, 1961. The address is built almost entirely on isocolon — parallel clauses of equal length and structure. See also: Gorgias of Leontini. Encomium of Helen. c. 414 BCE. Trans. Brian R. Donovan. The earliest surviving demonstration of isocolon as a deliberate rhetorical technique. Kennedy's inaugural and Gorgias's Encomium are separated by approximately 2,375 years. Both use the same device AI detectors flag.
29. ↑ Hemingway, Ernest. Prose style documented across the major novels and short stories. Hemingway's deliberately uniform sentence length — compar — is perhaps the most famous stylistic choice in twentieth-century American literature. The same technique governs legal prose: every appellate brief, every judicial opinion, every contract clause employs measured, equal-length clauses for authoritative rhythm. Caesar wrote this way. AI detectors flag this way.
30. ↑ Aristotle. Prior Analytics and Posterior Analytics. c. 350 BCE. Trans. A. J. Jenkinson & G. R. G. Mure. Available at MIT Classics. The foundational texts of formal logic. Source for the syllogism — the “Premise. Premise. Conclusion.” structure that AI detectors flag as “formulaic AI pattern.” Aristotle formalized deductive reasoning. Calling it a machine artifact is calling the foundation of Western logic a chatbot invention.
31. ↑ Etymonline.com. “delve (v.).” See also: Oxford English Dictionary. “delve, v.” Old English delfan, “to dig.” Documented before 1150 AD. The figurative sense — “to carry on laborious or continued research” — has been in continuous use since the mid-fifteenth century, approximately 573 years before ChatGPT. The word is the most frequently cited “AI tell” — and predates the technology by half a millennium.
32. ↑ Kobak, Dmitry et al. “Delving into ChatGPT usage in academic writing through excess vocabulary.” arXiv preprint arXiv:2406.07016 (2024). Identified statistically significant increases in the frequency of certain words (including “delve”) in academic papers published after the release of ChatGPT — but critically, all identified words existed in academic writing before LLMs. The study demonstrates frequency shift, not invention.
33. ↑ WriteHuman. “The Real Signature of AI Writing Isn't the Em-Dash Anymore.” 2026. Analysis of evolving AI detection heuristics and the shifting landscape of so-called “AI tells” — documenting that the markers detectors target change as models evolve, further undermining the premise that any stable set of linguistic features reliably indicates machine authorship.
34. ↑ The Economist. “AI Writing Comparison Analysis.” 2026. Systematic comparison of AI-generated and human-generated prose across multiple genres, demonstrating substantial stylistic overlap and the unreliability of surface-level feature detection as a classifier.
35. ↑ University of Pennsylvania, Department of English. “Joan Didion.” Critical essay. Penn Arts & Sciences. Analysis of Didion's prose style, including her characteristic use of em dashes, short declarative sentences, and repetitive structural patterns — the same features that AI detectors now flag as machine tells. Literary critics have long noted her “overuse of certain signature techniques” as a hallmark, not a deficiency.
36. ↑ Despey, Daniel. “Joan Didion's memoirs: substance & style.” a/b: Auto/Biography Studies 32.2 (2017): 349–351. Scholarly analysis of how Didion's stylistic repetitions and structural patterns function as deliberate literary devices rather than limitations — relevant to the article's argument that pattern consistency in prose is a feature of mastery, not evidence of automation.
37. ↑ Amis, Martin. “Joan Didion's Style.” London Review of Books 2.2 (1980). Extended critical analysis of Didion's prose mechanics — her em dashes, her sentence fragments, her controlled repetition. Amis documents precisely the techniques that AI detectors now catalog as machine-generated patterns, and attributes them to deliberate authorial control exercised decades before any language model existed.
38. ↑ Didion, Joan. “Why I Write.” New York Times Book Review. December 5, 1976. The author's own account of her compositional method: “Grammar is a piano I play by ear... The arrangement of the words matters, and the arrangement you want can be found in the picture in your mind.” Demonstrates that the stylistic patterns AI detectors flag are the product of conscious artistic decision-making.
39. ↑ Blair, Hugh. Lectures on Rhetoric and Belles Lettres. 1783. The standard English-language rhetoric textbook for over a century. Source for the transitional devices (“moreover,” “furthermore,” “in addition”) that AI detectors flag as machine-generated connectives. Blair taught these as fundamental elements of clear expository prose. Every composition course since has followed his model.
40. ↑ Virginia Museum of History & Culture. “John Marshall: Definer of a Nation.” See also: HISTORY.com. “John Marshall.” See also: Robarge, David. “John Marshall: The Formation of a Jurist.” St. John's Law Review 42.4 (1968). Marshall had one year of formal schooling and approximately six weeks of law lectures under George Wythe at William & Mary. He became the most consequential Chief Justice in American history and established judicial review in Marbury v. Madison (1803).
41. ↑ Lincoln's education is documented across: Herndon, William H. & Jesse W. Weik. Herndon's Lincoln. 1889. Repr. University of Illinois Press, 2006. See also: Donald, David Herbert. Lincoln. Simon & Schuster, 1995. Lincoln had approximately one year of aggregate formal education. He taught himself law by reading Blackstone's Commentaries on the Laws of England. He taught himself trigonometry from textbooks. He became President of the United States.
42. ↑ CNBC. “10 ultra-successful millionaire and billionaire college dropouts.” 2017. See also: Jobs, Steve. Stanford University Commencement Address. June 12, 2005. Available at Stanford News. Jobs dropped out of Reed College after one semester. He credited a calligraphy course — not a degree — for the typography that defined the Macintosh. Gates dropped out of Harvard after two years. Harvard calls him its “most successful dropout.”
43. ↑ Wikipedia. “List of autodidacts.” Cross-referenced with primary biographical sources for each individual cited. Faulkner dropped out of the University of Mississippi and won the Nobel Prize for Literature. Twain left school at twelve. Dickens left school at twelve. Whitman left school at eleven. Bradbury never attended college. Saramago never completed secondary school and won the Nobel Prize for Literature. Hume dropped out and became one of the most important philosophers in Western history. Harlan Ellison was expelled from Ohio State for hitting a professor who criticized his writing — and for twenty years sent that professor a copy of every work he published.
44. ↑ Carver, George Washington. Attributed. “If you love something enough, it will reveal itself to you.” Widely cited across biographical and educational sources. The statement captures the epistemological principle that sustained engagement with a subject produces knowledge that formal credentialing cannot substitute — the distinction between aspiration and inspiration that Section VII of this article develops.
45. ↑ Anthropic. “How Claude's text watermark works.” 2026. anthropic.com. Technical documentation of Anthropic's statistical text watermarking system for Claude. Describes how the watermark operates as a subtle statistical signal embedded in word choices — detectable in aggregate through Anthropic's proprietary detection tools but designed not to degrade output quality.
46. ↑ Anthropic Help Center. “How Claude marks AI-generated content.” 2026. support.anthropic.com. Consumer-facing documentation explaining Claude's content marking practices across text and images, including the conditions under which watermarks are applied and the limitations of detection when text is subsequently edited or incorporated into larger works.
47. ↑ TechCrunch. “Anthropic shares more details about how Claude's new watermarks will work.” 2026. Technical journalism reporting on the implementation details, reliability constraints, and privacy implications of Anthropic's watermarking system — including the fact that substantial human editing can degrade or eliminate the statistical watermark signal.
48. ↑ The Verge. “Claude will apply invisible watermarks to AI text and images.” 2026. Coverage of Anthropic's announcement of text and image watermarking capabilities for Claude, including C2PA-standard provenance metadata for images and the statistical text watermark for prose output.
49. ↑ Corbett, Edward P. J. Classical Rhetoric for the Modern Student. Oxford University Press, 1965. 4th ed. (with Robert J. Connors), 1999. The most widely used modern rhetoric textbook in American higher education. Systematically catalogs every rhetorical device discussed in Section IV of this article and teaches them as essential tools of effective argument and persuasion.

Article updated on 25 August 2026 at 4:46 PM PDT.